Overview
Prediction Tendency controls how selective the Prediction Map is when identifying prospective areas. It determines how closely new areas must resemble your positive Learning Data to receive a high VPS.
After completing Step 4: Select Input Features, you can switch between three prediction tendencies — Balanced (default), Conservative, and Aggressive — using the dropdown in the Prediction Results panel. Each setting produces its own Prediction Map, VPS results, Prediction Accuracy, Depth Accuracy, and Feature Importance list. You can create targets and export results from any or all three tendencies.
Note: If you do not see the dropdown in the Prediction Results panel, your experiment was created before this feature was released. Rerun from Step 4: Select Input Features to access it.
How to View Prediction Tendency Results
Complete Step 4: Select Input Features and open your prediction results.
Locate the dropdown in the Prediction Results panel. The default selection is Balanced.
Switch between Conservative, Balanced, and Aggressive at any time to compare results.
Review the Prediction Map, VPS results, Prediction Accuracy, Depth Accuracy, and Feature Importance list for each tendency.
Note: Each prediction tendency produces its own uncertainty, but DORA surfaces only the Balanced uncertainty. The Balanced tendency is optimized for the best overall accuracy, so its uncertainty is the most directly meaningful for exploration decisions. See Using the VPS x Uncertainty Filter and the Prediction Uncertainty Layer.
See the sections below for guidance on when to use each tendency.
Balanced Prediction Tendency
The Balanced tendency is suitable for most use cases. It maintains a middle ground between target confidence and spatial coverage.
Conservative Prediction Tendency
The Conservative tendency focuses on areas that closely match your known mineralization. It produces fewer targets with higher confidence, which reduces noise and limits false positives.
Consider using Conservative when:
Drill budgets are limited
You need to prioritize high-confidence targets for follow-up work
You are refining results from a previous experiment
Aggressive Prediction Tendency
The Aggressive tendency expands the range of highlighted areas. It includes targets that are less similar to your known mineralization and increases overall spatial coverage.
Consider using Aggressive when:
You are working in early-stage or grassroots exploration
You are screening a large or under-explored area
You are looking for unconventional or outlier targets
Create Targets
You can create targets from any or all three tendencies. Select a a tendency from the dropdown, then go to Create Targets to generate targets based on that result. Switch to another tendency and repeat as needed.
See Create Target Groups.
Export Results
All prediction results are available for export. Use Full Results to download outputs for all three tendencies, or select a specific tendency to download individually.
Additional Notes
Prediction Tendency options assume mineralized is the minority class in your Learning Data, which is the case in most experiments. If your mineralized (positive) points outnumber your non-mineralized (negative) points, the results may appear inverted: The Aggressive tendency will produce a sparser map than expected, while Conservative may produce broader coverage. If you notice this pattern, check the balance of your Learning Data in Step 2.
Learn More
Still Have Questions?
Reach out to your dedicated DORA contact or email support@vrify.com for more information.